Artificial Intelligence - Chapter 1 Introduction

人工智能 - 序

关于人工智能，主要是两个维度的考虑：thinking & acting, 对于每个维度， 我们又可以考虑他们以何种方式执行，即 humanly or rantionaly。 这样就产生了四种组合：

1. Acting humanly

What is Turing test: a computer or a system is put behind the screen and a human being is put behind another screen. You interact with one of them and try to figure out whether the 'computer' behind the screen is actually human or a computer.

The point is -- can a computer mimic or present itself as a human so strongly and so convincingly that you cannot tell the difference between a computer and a human.

Example application areas of AI:

      Natural language processing -- computer would listen and understand like human.

      Machine learning -- learn like human.

      Computer vision -- see like human.

      Robotics -- move like human. 

Aim: Build a system that does all these things that human can do and perhaps even more (go beyond capabilities of humans -- like alpha-go) 

In fact, most of the modern applications of Ai were achieved not necessarily by understanding human thinking. For example, deep learning has found a range of successes in the past decade and that's all based on artificial neural networks.

So, thinking humanly is probably sort of outdated definition or view of what AI is.

Thinking rationally is not really using. Why?

If we can figure out and write down all the rules of the world, then we can just use the power of logic and inference to help connect observations with understanding and predictions. That is, we make an observation, we appeal to our rule table or we appeal to our laws of thought that is written down in a giant book somewhere, then we can predict what would happen from one observation, or we can predict what the observation yields in the world.

But this is obviously infeasible, just think about how many rules can we really infer? How much can we really understand? We're still actually trying to understand all the business today!

So thinking rationally also sort of ran into some limitations and that definition also became pallidated at some point.

Acting rationally is gonna be the one we are going to be using in terms of the modern definition in the modern perspective of AI. Modern AI is based on the notion of acting rationally.

Rationality means t